36
Forget Big Data - Think Machine Learning! Jane Zavalishina, CEO of Yandex Data Factory

Big Data Week: Forget Big Data - Think Machine Learning!

Embed Size (px)

Citation preview

Forget Big Data - Think Machine Learning!

Jane Zavalishina, CEO of Yandex Data Factory

Big Data is a new oil

Finextra, 09.01.2015 on research from Aite Group

│ …Big data is a particular sore │ point, invoking dissatisfaction │ among 76% of North │ American bankers…

Why the bankers are so unhappy?

Volume More data than ever

Velocity High speed of change

Variety Different types, a lot of unstructured data

problemBig data = Big

problemBig data = Big IT

projectsBig data = Big IT

budgetsBig data = Big IT

What about value?

Here’s how most people imagine use of Big Data

Here’s how it will actually look like

Getting insights to help you make decisions

Brings only a fraction of value

The true economic value of Big Data

Using machine learning to automate decision making

It’s true that change is coming (and data are generated) soquickly that human-in-the-loopinvolvement in all decision making is rapidly  becoming  impractical. Looking three to five years out, we expect to see far higher levels of artificial intelligence...  

McKinsey & Company, “An executive’s guide to machine learning”, June 2015

Robots are causing a new Industrial Revolution. Similar to what happened to farming, 70 percent (or more) of current jobs will be replaced by machines. Replacement by robots in most jobs is just a matter of time.  

American Thinker, “The Next Phase of the Industrial  Revolution”, June 2015

The robots might be coming for your job, even if you think it seems safe.

Business Insider, “Jobs replaced by robots”, June 2015

Deep Blue

Self driving car

Author: smoothgroover22 by CC BY 4.0 https://clck.ru/9YibV

21

8

54,548

53,966

51,108

47,815

28,152

25,900

25,185

22,940

19,043

16,751

Top 10 Newspapers by Digital Traffic

USAToday.com

Total number of unique visitors for January 2015 (in thousands)

NYTimes.com

DailyMail.co.uk

WashingtonPost.com

TheGardian.com

NYDailynews.com

LATimes.com

NYPost.com

SFGate.com

Telegraph.co.uk

1

2

3

4

5

6

7

9

10

8

54,548

53,966

51,108

47,815

28,152

25,900

25,185

22,940

19,043

Top 10 Newspapers by Digital Traffic

USAToday.com

Total number of unique visitors for January 2015 (in thousands)

NYTimes.com

DailyMail.co.uk

WashingtonPost.com

TheGardian.com

NYDailynews.com

LATimes.com

NYPost.com

SFGate.com

Yandex.News

1

2

3

4

5

6

7

9

10

23,164

23 000 000 monthly readers

0 editorial team

Online Gaming: Loyalty Management & Personalisation

Predicting & Preventing User Churn

Gamer data (number of victories, battles, purchase logs etc.) External data (e.g. weather)

Indication of potential churner Recommendation of the best retention offer

Retail Business: Next Best Offer

Upsell recommendations for a Bank

Customer profiles Historical data on communications and responses

13% increase in NPV

Anomaly detection for CERN (LHCb)

Up to 30 mln collisions per second terabytes of data per second several thousand of parameters to check

Industry and infrastructure: Predictive maintenance

© IBM

“It’s estimated that 90 percent of the data in the world today has been created in the last two years alone.”

Data quickly becomes commodity

Here’s our prediction

In all the business processes, where:

• we know exactly what we want to improve and can measure it

• we have enough data • we can experiment • we can take automated action

In 10 years, we’ll have algorithms doing the work.

What’s your plan?

Not big data strategy, but ML strategyBig data by itself means costs, not value

Value first: start with a few short projects todayThere’s no value without implementation

Experiment and measurement is the key Continuous experimenting is the only way to stay on top

32

McKinsey & Company, “An executive’s guide to machine learning”, June 2015

...Because machine learning’s emergence  as a mainstream management tool is relatively recent, it often raises questions

Jane Zavalishina

CEO Yandex Data Factory

Happy to answer your questions!

[email protected]

Yandex Data Factory

Created in 2014

Apply Yandex’s machine learning expertise to other industries

Computational infrastructure

Proprietary machine learning tools

Data scientists

Title: Open Sans 100 px

• Subtitle:OpenSans48px